CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   Timestep with mesh deformation (

rbarrett July 19, 2011 11:44

Timestep with mesh deformation
Hi all,

I am attempting to simulate the compression of a piston within a cylinder. The model is axisymmetric and I have used Ansys Meshing to generate the mesh. Mesh deformation is employed through the use of CEL expressions, coupled together with the use of 'Specified Displacement' mesh motion.

The problem that I am having is related to convergence 11.6 ms into a 16.6 ms compression time. Up to this point, relatively good convergence is obtained and then the solution diverges. I have tried reducing the size of the time step, but reducing this value beyond a certain point (3e-04 [s]), the model fails due to a negative volume error at multiple locations throughout the deforming zone.

I would greatly appreciate any help on this matter, particularly in relation to the size of the timestep employed.

ghorrocks July 19, 2011 19:22

Highly compressing a mesh using moving mesh is difficult, you frequently get negative volume errors. You might be able to improve things using mesh weighting factors, but the biggest step forwards in this area is in CFX V14 where you have much more control over the mesh motion diffusion.

rbarrett July 21, 2011 11:40

Hi ghorrocks. Thanks for your reply. Unfortunately I have only access to Ansys v12.1. Are there any other workarounds within CFX itself that may allow for more accurate modeling of a large compression, i.e. inputting a new mesh and interpolating a results file to give the initial values?

stumpy July 21, 2011 14:49

Glenn, out of interest, what's in v14 that's new for mesh motion diffusion?

ghorrocks July 21, 2011 19:25

For the V14 new stuff check out the ANSYS Community portal. One thing which caught my eye in this area is the ability to define an anisotropic mesh diffusion for the mesh smoothing step. The demo case using it compared a mesh being compressed which distorted using normal isotropic diffusion, but the anisotropic diffusion kept the mesh nice as the mesh was squashed.

Richard - if you want to model a large compression and want to stay on V12 then you have two options: Either use fortran to define the mesh motion, and then you can control it directly; or put a remeshing step in there.

All times are GMT -4. The time now is 15:57.